Background
Most patients with epithelial ovarian carcinoma (EOC) are diagnosed with stage III or IV disease, i.e., when dissemination is already at hand. Along with rapid development of chemoresistant disease, this brings down the 5-year disease-specific survival in EOC to 50% [
1]. Dissemination is mainly abdominal via tumor cells in ascites. These are likely enriched for aggressive cells that may have exfoliated not only from an excised primary tumor but also from established and often numerous metastases.
Ascitic ovarian cancer cells are proposed to undergo epithelial-to-mesenchymal-transition (EMT) to a motile phenotype with low levels of E-cadherin and higher invasivity than the primary tumor cells [
2,
3]. They are thus similar to blood- or lymph-borne circulating tumor cells from other types of carcinoma. They are similar also in that they represent a less studied but phenotypically and pathologically progressed stage compared to the primary tumors.
Of importance for the present study, ascitic tumor cells may occur as single-cell suspensions and sheet-like cell aggregates, but occasionally also as compact spheres in which individual cells cannot be discerned. Although spheres have been associated with invasivity, e.g., in
in vitro models using artificially created spheroids from cell lines [
4‐
6], these
in vivo populations have not been extensively characterized. Interestingly,”collective migration” of multicellular clusters of cancer cells has been described as an enhanced type of metastatic behavior in breast and colon cancer [
7].
Sphere formation
in vitro, dissemination and chemoresistance are in turn features associated with cancer stem cells or tumor-initiating cells (TICs). TICs are now regarded as a major cause of relapse in chemoresistant cancer, and thus need to be identified both for prognostic and therapeutic purposes [
8‐
10]. In the experimental
in vitro setting, TIC identity is commonly based on expression of two or three of a set of markers, and is verified based on high tumorigenicity and/or serial transplantation in nude mice. Irrespective of whether TICs originate from stem cells or from a certain subpopulation of cancer cells, they are regulated not only by genetic but also by epigenetic and adaptational events[
11]. As part of adaptation to hypoxia, both TICs and normal stem cells depend on glycolysis rather than mitochondrial respiration for ATP. Importantly, hypoxia has been shown to promote tumor”stemness” as well as EMT [
12,
13], suggesting that in ascitic EOC spheres, the hypoxic interior might enrich for TICs.
Cancer-associated fibroblasts (CAFs) represent yet another parameter contributing to invasivity, metastasis and chemoresistance [
14]. Immunohistochemical analysis has shown absence of the CAF marker smooth muscle actin (α-SMA) in normal ovarian tissue, whereas the stroma of metastatic EOC tumors abounded in CAFs [
15].
To explore the molecular pathology of EOC ascitic cell populations, we have here examined spheres and single-cell populations that were isolated directly from patients. The samples were compared in terms of expression of metabolic, EMT, TIC and CAF markers, and analysed by western blot in order to specifically assess the net outcome of genetic, epigenetic and post-translational effects. Importantly, freshly isolated human EOC ascitic samples were used rather than subcultured clones or cell lines, in which pheno- and genotypic drift is a wellknown problem. As metabolic marker we used mitochondrial β-F1-ATPsynthase, which reflects mitochondrial oxidative respiration and is downregulated at the mRNA level in cancer cells [
16]. It has prognostic power on its own and in a ratio to Hsp60, i.e., as a bioenergetic cell (BEC) index shown to be indicative of low mitochondrial respiration and worse prognosis in, e.g., breast, lung and head-and-neck cancer [
17,
18]. We also examined the mitochondrial transcription factor A, TFAM, as it correlates with cellular mitochondrial content [
19,
20] and hence indirectly with oxidative metabolism. With regard to EMT and motility, we examined E-cadherin and vimentin as standard markers of epithelial and mesenchymal cells, respectively [
7]. For CAF markers, we used α-SMA and the receptor for PDGFβ (PDGFβR) [
14,
15].
The significance and use of TIC markers requires some consideration. For cell sorting/isolation purposes, the surface proteins CD44, CD117 [
10,
21,
22] and EpCAM [
23,
24] are convenient antibody targets. They also have intriguing dual functions as adhesion molecules as well as receptors [
11]. In TICs isolated from primary breast cancer, shRNA-mediated knockdown of CD44 led to decreased expression of stemness-associated genes and to loss of tumorigenicity [
25]. CD133 is another common TIC marker, including in EOC [
26]. However, CD133 has been questioned as a marker of tumor-initiating capability in EOC [
22,
27], and there is uncertainty as to the roles of different splice and glycosylation forms and different antibody clones [
28,
29].
TICs are also characterized by transcription factors that are essential for stem cell self-renewal and pluripotency, and which include Nanog and Oct-4A [
10,
11,
22]. Several studies have used combinations of surface and intracellular markers, e.g., CD44, CD117, Nanog and Oct-4 to characterize TICs derived from experimental EOC spheroids created
in vitro[
10]. The same report also used ABCG2, the membrane efflux pump which defines a side population enriched for TICs [
30‐
32] and which contributes to chemotherapy resistance [
33].
In summary, it is clear that the biology and molecular pathology of malignant ascites are of importance for the understanding and treatment of this deadly disease [
6]. There is, however, still a dearth of information on ascitic cell subpopulations
in vivo, i.e., isolated directly from patients. Based on such material and on protein marker profiles reflecting known phenotypic traits, our results indicate potentially very high levels of circulating TIC-type cells in malignant ascites and demonstrate that ascitic spheres and monolayer-forming cells are two distinct
in vivo populations with different tumorigenic potential, based on protein markers. For the first time, the presence in malignant ascites of cancer-associated fibroblasts is also indicated.
Discussion
Compared to experimental studies on tumor-initiating cells (TICs), the representation of TICs in the actual human in vivo situation is less well studied. This fact, and the nature of ovarian cancer ascitic cells as circulating and potentially metastatic cells, prompted us to examine freshly isolated ascitic samples for expression markers of EMT, TICs and cancer-associated fibroblasts (CAFs).
Previous studies have suggested that EOC ascitic cell spheres represent the invasive and/or metastasis-forming subpopulation leading to recurrent disease [
4,
6]. This notion has in part been supported by
in vitro work on artificial spheroids, and is also based on the well-known fact that TICs can be isolated from
in vitro spheroids cultured continuously in stem cell medium. Indeed, our own initial hypothesis was that the spontaneous EOC ascitic spheres are similar to TIC spheroids and that the hypoxic interiors of spheres would support or harbor TIC-like cells.
However, our results show that the S-type populations freshly isolated from patients were not only of an epithelial and thus less invasive phenotype, but they were also low in the EOC TIC marker CD44 and the stem cell transcription factor Oct-4A and did not express CAF markers. Compared to the M-type samples, the S-type thus present a less tumorigenic profile. It may be noted, however, that they likely represent a chemoresistant population since chemotherapeutic drugs do not penetrate such multicellular structures [
34,
43].
Among M-type samples, two subtypes could be discerned: one that was similar to S-type but obviously lacking some factor required for sphere formation in ascites, and a second subtype that was CD44
high/Oct-4A
high and which in addition also expressed specific CAF markers. In addition to other reports on CD44 as an EOC TIC marker [
10,
21,
22], the role of CD44 in regulating a TIC phenotype was recently shown when shRNA-mediated knockdown of CD44 caused breast cancer TICs to differentiate and abolished their tumorigenicity in mice[
25]. Similarly, results of experimental knockdown of Oct-4 have indicated its role in regulation of a TIC phenotype [
44].
We therefore propose that the second, CD44high/Oct-4Ahigh subtype of M is the more malignant one. This is based on the roles of CD44 and Oct-4A in defining TICs, on the motility/invasivity of vimentin-rich tumor cells, and on the expression of CAF markers α-SMA and PDGFβ-R.
In further support of this model, the inverse correlation which we observed between CD44 and E-cadherin may reflect the recent finding that in serial xenografts of EOC tumors, the tumorigenic CD44
high cells were low/intermediate in E-cadherin [
45]. Moreover, CAFs may originate either as fibroblasts or from epithelial cells that have undergone EMT [
14]. Here, the ratio of E-cadherin:vimentin in M-type samples and the adherence-independence of these cells
in vivo are both in accordance with the latter scenario. Importantly, the observed co-expression of PDGFβ-R and α-SMA further strengthens the identity and presence of activated CAFs, whose role(s) as direct promoters of metastasis, angiogenesis and chemoresistance is increasingly clear[
14,
15,
46].
In line with such roles, we found that samples with a diagnosis of clinical stage IV showed higher expression of α-SMA than did samples from stage IIIC patients (
p = 0.040). This is in line with the recent finding that CAFs promote EOC growth and metastasis
in vivo[
15]. It is also in line with stage IV being defined by hematogenous metastasis to distant sites, e.g., to the liver. Altogether, this is to our knowledge the first report to indicate the presence of CAFs in EOC malignant ascites.
Transcription factors generally make up only a minute fraction of the entire cellular proteome, and TIC-associated transcription factors an even smaller fraction within a population. Yet we could easily detect Oct-4A and Nanog in the present material, indicating significant levels of TICs within the samples. It is likely of great importance for this result that the samples were not subcultured in vitro, where the differentiating conditions and the consequent asymmetric proliferation of TICs will soon dilute the levels of TIC markers.
We note that although Oct-4 was recently associated with advanced FIGO stage and higher histological grade in serous ovarian adenocarcinoma [
47], the present report is the first to show high expression in EOC of Oct-4A, the isoform that is crucial for stem cell renewal[
39]. We are hence also first to show high expression in EOC of the stabilized, and more active, SUMOylated form [
48] of Oct-4A. Indeed, 24/27 (89%) of the samples expressed Oct-4A/SUMO-Oct4A. Moreover, and of importance for data interpretation in further studies, we found that in some samples virtually all Oct-4A was in the ~90 kDa SUMOylated form. This demonstrates that lack of ~40 kDa Oct-4A in a western blot does not necessarily correspond to lack of the protein. Interestingly, Ubc9 is the only SUMO-ligase required for SUMO-Oct-4A [
48], and it is upregulated in ovarian cancer [
49], although its prognostic potential has not been investigated.
The 100 K is likely the calpain-cleaved 100 kDa E-cadherin fragment that releases β-catenin allowing it to translocate to the nucleus and participate in Wnt-like signaling [
41,
50]. The antibody used here does not recognize any other known intracellularly cleaved fragment. Nor could the fragment be due to the well-studied extracellular cleavage of E-cadherin that generates the soluble extracellular form, sometimes known as gp80 or sE-cad, which would not be found in cell lysates nor be recognized by the antibody used. Moreover, the remaining cellular fragment from sE-cad is around 37 kDa, i.e., not the one we observed. It would be interesting to know whether 100 K is identical to the cytoplasmic E-cadherin observed at high levels in the multipotent, tumorigenic epithelial-mesenchymal hybrid subset of EOC tumors [
45], and/or to internalized E-cadherin leading to increased β-catenin/Wnt signaling in EOC cells [
51].
Like Oct-4A, Nanog is involved in stem cell renewal. In EOC, Nanog expression in tumors has been shown to correlate with clinical stage [
52]. Here, Nanog correlated with TIC marker EpCAM (
r
s
= 0.694,
p = <0.001) and with 100 K (
r
s
= 0.425,
p = 0.027), but not with CD44 or Oct-4A. The results thus show that TIC surface and transcriptional markers do not necessarily coincide. This in turn suggests that while abundant reports show that cell sorting based on CD44 will enrich for TICs, there may
in vivo exist several subsets of TICs. Applied to the present material, the model also suggests that the Nanog
high/EpCAM
high samples represent a particular TIC subset which may be either M- or S-type, and which is separate from the CD44
high/Oct-4A
high subset observed only in M-type samples.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MW did all the lab work and statistics and contributed to writing and revision. EH and EÅL contributed all clinical material, clinical data and medical expertise. MS conceived of the study and drafted the manuscript. All authors read and approved the final manuscript.